Detection of roadside vegetation using features from the visible spectrum

Abstract

Detection of vegetation in images is a common procedure in remote sensing and is commonly applied to satellite and aerial images. Recently it has been applied to images recorded from within ground vehicles for autonomous navigation in outdoor environments. In this paper we present a method for roadside vegetation detection intended for traffic safety and infrastructure maintenance. While many published methods for vegetation detection are using Near Infrared images which are particularly suitable for vegetation detection, our method uses image features from the visible spectrum allowing the use of common onboard color cameras. Our feature set consists of color features and texture features. One of our specific goals was to identify a useful texture feature set for the problem of vegetation detection. Based on the feature set, the detection is implemented using a Support Vector Machine algorithm. For training and testing purposes we recorded our own image database consisting of different images containing roadside vegetation in various conditions. We are presenting promising experimental results and a discussion of specific problems experienced or expected in real-world application of the method.

Superpixel-based class-semantic texton occurrences for natural roadside vegetation segmentation

“…A major benefit of visible features is that they retain high consistency with human visual perception. Color is one of the dominant resources that the human eyes depend on in the perception and discrimination of different objects, and it is one of most popular features in existing research on vegetation segmentation, which mainly focuses on investigating the suitability of different types of color spaces, including CIELab [4], YUV [5], HSV [6], and RGB [7] etc. Although vegetation is widely known as being characterized primarily by a green or orange color, it is still a challenge to find a suitable color representation of vegetation in complex natural conditions.…”

“…However, it may not be the case in scenes containing sky and with varying lighting conditions, such as the presence of shadow, shining, under-and overexposure effects. Another popular type of visible features is texture, which is often represented by performing wavelet filters, such as Gabor filters [8] and Continuous Wavelet Transform (CWT) [6], extracting pixel intensity distributions, such as pixel intensity differences (PIDs) [4], [5] and variations in a neighborhood [9], [10], or generating spatial statistic measures [10], entropy [7], or statistical features over superpixels [11]. Table 1 lists typical visible approaches for vegetation segmentation in existing studies.…”

“…One of the early studies on vegetation segmentation in outdoor images was presented in [12], which employed a self-organising map (SOM) for object segmentation and then extracted color, texture, shape, size, centroid and contextual features of segmented regions for 11 object classification using a multi-layer perceptron (MLP). In [7], the entropy was used as a texture feature, together with RGB color components and an SVM classifier for detecting vegetation from roadside images. The intensity differences between pixels were combined with a 3D Gaussian model of YUV channels for grass detection [5], and with L, a, and b components for object segmentation [4].…”

Motion estimation aided detection of roadside vegetation

“…mowing the grass along the road). The method presented in this paper is a continuation of our previous work [9] where we used a simpler set of features and a traditional approach of machine learning and classification. This research has been partially supported by the European Union from the European Regional Development Fund by the project IPA2007/HR/16IPO/001-040514 "VISTA -Computer Vision Innovations for Safe Traffic".…”

Analysis of Hybrid Classification Approach to Differentiate Dense and Non-dense Grass Regions

“…Meanwhile several approaches on weed classification have been developed using various classifiers such as k-Nearest Neighbour [2], Adaptive boost [3], Artificial Neural Network [4], Support Vector Machine [5] and Wavelet [6]. The increased use of automation on roadside data is the main reason for motivating into the research on detection and classification of vegetation region from roadside [7] [8] [9].…”